#Signing in to Claude & Codex
There are two kinds of sign-in in Chatspaces, and it's worth keeping them straight:
- Sign in to Chatspaces with GitHub — once, when you first open the app. This is how Chatspaces knows which repos you can work on.
- Sign in to your agent (Claude or Codex) — the first time you use it. This connects your own Claude or ChatGPT account so the agent can do work on your plan.
This page is about the second one.
#What you need
You use Chatspaces with your own Claude or ChatGPT account — Chatspaces doesn't provide the agent's model access or an API key. The agent runs on your machine and is driven by your account, so what you need depends on which agent you pick:
| Agent | Sign in with | What's required |
|---|---|---|
| Claude | Your Claude (Anthropic) account | A paid plan — Claude Pro or Claude Max. A free Claude account won't work. |
| Codex | Your ChatGPT (OpenAI) account | Any ChatGPT account — the free plan works. Paid plans (Plus/Pro) just give you more headroom. |
In short: Claude needs a paid subscription; Codex works on a free ChatGPT account.
Usage counts against your own Claude or ChatGPT account. Because Chatspaces drives the agent through your account (the same as if you ran it on your laptop), the model usage — and any rate limits or quotas — are yours, on your Claude/ChatGPT plan. Chatspaces separately meters the machine it runs on (CPU and memory); see Usage & limits. They're two different meters: your agent plan covers the thinking, your Chatspaces plan covers the box.
#How sign-in works
The first time you open a chatspace with a given agent, it asks you to authenticate. You never type a password into Chatspaces — it's the agent's own secure login flow.
#Claude
The chatspace shows a login link. Open it, approve access in your browser, and the chatspace connects. That's it.
#Codex
The chatspace shows a short device code and a link (to OpenAI's device-login page). Open the link, enter the code, and the chatspace finishes connecting on its own — nothing to paste back.
Codex sign-in requires device code login to be enabled on your ChatGPT account. OpenAI turns this off by default, and without it your code is rejected. Enable it once at ChatGPT → Settings → Security → "Allow device code login". On a ChatGPT workspace/Enterprise account, a workspace admin enables it under Workspace Settings → Permissions instead.
The Codex device code is short-lived (it expires after a few minutes). If it times out, the chatspace will show a fresh one — just try again.
#Staying signed in
Re-authenticating on every new chatspace would be tedious, so Chatspaces can securely remember your agent login and reuse it on future chatspaces. With it saved, new chatspaces skip the sign-in step entirely.
This is optional and protected carefully — your credentials are encrypted, never shown back to you, and scrubbed from logs. You can opt out, in which case you'll sign in each time. The full story, including how to opt out, is in Provider credentials.
#When you'll be asked to sign in again
- You opted out of saving your login — you'll sign in each session.
- Your saved login expired or was revoked (for example, you signed out of Claude/ChatGPT elsewhere, or your plan lapsed). The chatspace will surface the login flow again; sign in and it'll save the fresh login (if you allow it).
#Switching agents
You choose the agent per chatspace, so you can run some chatspaces on Claude and others on Codex. Each agent has its own sign-in; signing in to one doesn't affect the other.